j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8

H O S T E D BY Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/jtte

Original Research Paper

Application of semi-analytical finite element

method coupled with infinite element for analysis
of asphalt pavement structural response

Pengfei Liu*, Dawei Wang, Markus Oeser

Faculty of Civil Engineering, RWTH Aachen University, Aachen D-52074, Germany

article info abstract

Article history: A specific computational program SAFEM was developed based on semi-analytical finite
Available online 20 January 2015 element (FE) method for analysis of asphalt pavement structural responses under static
loads. The reliability and efficiency of this FE program was proved by comparison with the
Keywords: general commercial FE software ABAQUS. In order to further reduce the computational
Asphalt pavement time without decrease of the accuracy, the infinite element was added to this program. The
Structural response results of the finite-infinite element coupling analysis were compared with those of finite
Semi-analytical finite element element analysis derived from the verified FE program. The study shows that finite-infinite
method element coupling analysis has higher reliability and efficiency.
ABAQUS 2015 Periodical Offices of Chang'an University. Production and hosting by Elsevier B.V. on
Infinite element behalf of Owner. This is an open access article under the CC BY-NC-ND license (http://
Finite-infinite element coupling creativecommons.org/licenses/by-nc-nd/4.0/).
analysis

For a typical pavement structure problem as shown in

1. Introduction
In the past decades, the finite element (FE) method has been
developed rapidly and was increasingly used in many indus-
trial fields as well as in the routine pavement design process.
But there are several limitations in the conventional FE
packages, such as the complexity of the program and hence
the time-consuming user training process or over-
simplifications of the modeling (Liu et al., 2014a). Therefore, it
is necessary to find means that both improve the computa-
tional speed without increasing the resource requirement and
keep the computational accuracy. One specified program
SAFEM was developed based on the semi-analytical FE
method to meet these requirements.
Fig. 1, the geometry and material properties usually do not
vary in the z-direction, but the boundary conditions, e.g. the
load terms, exhibit a significant variation in that direction.
Due to this characteristic, the pavement structure problem
could not be simplified as a 2D plane strain case. An alterna-
tive method is to simplify the pavement structure as a 2D
axisymmetric case, and the response under multiple loads
can be obtained using superposition principle. However, the
models based on the axisymmetric formulation cannot real-
istically model unidirectional loads, cracks or discontinuities
within a pavement system and the ability to realistically
model non-uniform contact pressures is also limited (Fritz,
2002). Zienkiewicz and Taylor (2005) proposed one method
that assuming the displacements in the geometrical z-

* Corresponding author. Tel.: 49 241 8020389; fax: 49 241 80 22141.

E-mail address: liu@isac.rwth-aachen.de (P. Liu).
Peer review under responsibility of Periodical Offices of Chang'an University.
http://dx.doi.org/10.1016/j.jtte.2015.01.005
2095-7564/ 2015 Periodical Offices of Chang'an University. Production and hosting by Elsevier B.V. on behalf of Owner. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8 49

2. Description of semi-analytical finite

element method

The first step in the FE formulation of SAFEM is to express the

element coordinates and element displacements in the form
of interpolation functions using the natural coordinate system
of the element.
By using the SAFEM, the general form of the shape func-
tions can be written as a Fourier series in which z ranges be-
tween 0 and a (Fritz, 2002; Hu et al., 2008; Liu et al., 2013, 2014a,
2014b; Zienkiewicz and Taylor, 2005), as shown in Fig. 2.
Fig. 1 e Pavement structure geometry and load mode (Liu X
6
et al., 2014a). u Nk x; y; zuk
k1
(  " #)
XL X 6
lpz lpz
Nk x; ycos Nk x; ysin ulk (1)
direction can be represented using Fourier series and l1 k1
a a
exploiting its orthogonal properties, the problem of such a
where l identifies the term of the Fourier series, L is the total
class can thus be simplified into a series of 2D solutions. This
number of Fourier terms considered, Nk and Nk are the shape
method is so called the semi-analytical FE method which has
functions of the element at node k.
potential for overcoming the difficulties mentioned above.
The loading function defining the variation of load along
The pavement length along the traffic direction and the
the z-direction is given by Zienkiewicz and Taylor (2005).
thickness of sub-grade can be considered infinity. In order to
minimize the influence of these boundaries, the FE-mesh XL    
lpz lpz
must be sufficiently big in horizontal and vertical directions. f px; ycos px; ysin (2)
l1
a a
This, however, means that a relatively large number of FE-
elements are required to appropriately discretize the mesh. where px; y and px; y represent the pavement load.
One of the methods that can keep the number of FE-elements The pavement is assumed to be supported on both side
reasonably low is the application of infinite elements at the faces (z 0 and z a) in a manner preventing all displace-
infinite boundaries of the system in the FE method (Li et al., ments in the xy plane but permitting unrestricted motion in
2007b). This concept of the infinite element was firstly pro- the z-direction. The Fourier series expansion should meet this
posed by Ungless (1973), and then modified and further requirement of the boundary condition. The displacement
developed by many other researchers such as Beer and Meer functions with three components u, v and w can be written as
(1981), Bettess (1977, 1980), Bettess and Zienkiewicz (1977), follow
Zienkiewicz et al. (1983) and applied to a variety of problems 2 3
lpz
(Hjelmstad et al., 1997; Li et al., 2007a; Wang and Brill, 2013). 6 sin a 0 0 78
8 9 6 7 9
Currently, the infinite elements have been developed from <u= X L X 6 6 7< ulk =
6 lpz 7
one-dimensional (1D), unidirectional to 3D and multidirec- U v Nk 6 0 sin 0 7 vlk
: ; 6 a 7: l ;
tional approaches. The concept of infinite elements can be w l1 k1 6 7 wk
4 lpz 5
generally divided into two categories: the mapping infinite 0 0 cos
a
element and decaying infinite element (Sallah and Buchanan, X
L
1990). The features of the infinite elements can be concluded Nl $Ul (3)
l1
as follows (Jiang et al., 2009; Zhu and Bian, 2001):

 The mapping from finite field in the natural coordinate to

infinite field in the global coordinate, e.g. when the natural
coordinate x approaches 1 the corresponding global coor-
dinate will trend towards infinity to fulfill the requirement
that the computational field trends towards infinity.
 The decay of the displacement in the infinite field, e.g.,
when x approaches 1, the displacement tends towards 0 to
fulfill the requirement of the boundary condition that the
displacement at the infinity is 0.

In the following sections, the mathematical basis of the

SAFEM and the 2D mapping infinite element will be presented at
first, followed by a verification of the SAFEM by comparison with
ABAQUS as well as a verification of the finite-infinite element
coupling analysis in the SAFEM program. Finally, a brief sum- Fig. 2 e Schematic representation of a SAFEM situation (Liu
mary and conclusions are provided at the end of this paper. et al., 2014a).
50 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8

where ulk , vlk and wlk are the displacements of the node at the
e 

l T l 
1 l T l
Fourier term k in x-, y- and z-directions, respectively. Kllgk a area Bg DBk Bg DBk dxdy l 1; 2/ (10)
2
Similarly, the loading function for the pavement analysis
can be formulated as follows (Fritz, 2002) where g and k represent the nodes of the element, respec-
tively, area is the area of the element.
X
L
lpz X L
A typical term for the force vector becomes
f px; ysin fpgl (4)
l1
a l1  l e  T
F Nl fpgl dxdydz (11)
n    vol
X 2Pt lp lp
px; y cos Zt1  cos Zt2 (5) The final assembled equations have the following form
t1
lp a a
2 38 1 9 8 1 9
where Pt is the tire load pressure, Zt1 is the z coordinate where K11 >
>U > > >>F >>
6 K 22 7< U2 = < F2 =
6 7
the tire load area starts, Zt2 is the z coordinate where the tire 4 1 5> > > > 0
(12)
> > > : L>
load area ends. LL : L ; ;
K U F
By using the principle of minimum potential energy, a
typical sub-matrix of the element stiffness matrix (Klm)e is (Hu Eq. (12) shows that the large system of equations splits up
et al., 2008) into L separate problems.

 lm e  T Kll Ul Fl 0 (13)
K vol Bl DBm dxdydz (6)
According to Eq. (13), the Fourier expansion of the loading
One detail worth to be mentioned is the decomposition of
factors involves only one term for a particular harmonic, so
the strain-displacement matrix Blk as follow
only one set of simultaneous equations needs to be solved.
2 3
vNk lpz This solution is just like a 2D plane problem. The sub-
6 vx sin 0 0 7
6 a 7 displacement vector calculated from each term of Fourier
6 7
6 vNk lpz 7 series only needs to be assembled to a global vector.
6 0 sin 0 7
6 vy a 7
6 7
6 7
6 lp lpz 7
6 0 0  N sin 7
6 a
k
a 7
6 7
Bk 6
l
7 3. Description of 2D mapping infinite
6 vNk lpz vNk lpz 7
6 7
6 vy sin a vx
sin
a
0 7 element method
6 7
6 7
6 lp lpz vNk lpz 7
6 0 Nk cos cos 7
6 a a vy a 7
Due to the 2D mesh of the finite elements used in the SAFEM,
6 7
6 7 only 2D mapping infinite elements which were applied in this
4 lp lpz vNk lpz 5
Nk cos 0 cos program are introduced in this section. The 2D mapping
a a vx a
infinite elements can be divided into two types which are
l lpz l lpz
Bk sin Bk cos (7) unidirectional and bidirectional ones.
a a
The Blk in Eq. (7) is splitted into two matrices of which each
only includes one type of trigonometric terms. 3.1. Formulation of 2D mapping infinite elements
From Eqs. (6) and (7), the stiffness matrix of one element
includes (Zienkiewicz and Taylor, 2005) Fig. 3(a) shows the 2D unidirectional infinite element which
Za Za extends to infinity in y-direction. The nodes 1, 2 and 3 are at
sinlpz cosmpz sinlpz sinmpz the interface which can be coupled with finite elements and
I1 $ dz; I2 $ dz;
a a a a the nodes 4 and 5 are in the middle of the infinite element.
0 0
(8)
Za Fig. 3(b) shows the bidirectional infinite element, in which the
coslpz cosmpz
I3 $ dz node 1 can be coupled with a finite element and nodes 2 and 3
a a
0 are in the middle of the infinite element. Through mapping
the infinite reference element is transferred from the global
The integrals exhibit orthogonal properties which ensure
coordinate to a parent element in a finite region, i.e. the
that
element in natural coordinate with 1  x  1 and1  h  1.
8
< 1 a for l m The transformation equations, or the mapping infinite
I2 I3 2 (9) element, between the global and natural coordinates are
:
0 for lsm (Zhao, 2012; Zhou et al., 2004)
Only when l and m are both odd or even numbers, the first X
n

integral I1 is zero. But due to the special structure of the Bl x Mi xi ;

i1
matrix, all terms that include I1 vanish (becoming 0). This
X  (14)
n
n 5 for unidirectional infinite element
means that the matrix (Klm)e becomes a diagonal one. In other y Mi yi
n3 for bidirectional infinite element
words, the non-zero values are only located in the diagonal i1

area where l m. Thus, the stiffness matrix can be reduced to P

where the node-wise mapping functions Mi 1, as shown in
(Zienkiewicz and Taylor, 2005) Table 1.
 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8 51

3.2. Finite-infinite element coupling method

When analyzing a problem with an infinite domain, the

infinite elements can be combined with the finite element
method, i.e. the near field is analyzed by finite elements and
the far field is simulated by infinite elements. As a result, the
problem with continuous infinite degrees of freedom can be
converted into one with discretized finite degrees of freedom.
The unknown displacement at any node in the computational
domain can be computed from the equilibrium equations,
which are derived from the equilibrium conditions and the
consistency of displacements at the interface between finite
and infinite elements. The general procedure of the finite-
infinite element coupling method is illustrated as follows
(Zhao, 2012).

 The equations are built up in the near and far field,

respectively.

  
L1 C1 ; F1 0
(16)
L2 C2 ; F2 0

where {C} is unknown variable, {F} is the interaction force at

the interface, 1 and 2 represent the finite element and infinite
element fields, respectively.

 Derivation of the displacement expression at the interface

Fig. 3 e 2D mapping infinite elements.
  
fU1 g f1 C1 ; F1
(17)
fU2 g f2 C2 ; F2

 Establishment of the continuous equations at the interface

The displacement of any point in the parent element can 
be expressed as follows (Zhao, 2012; Zhou et al., 2004) fU1 g fU2 g
(18)
fF1 g fF2 g f0g

X
n
u Ni ui ; The three equations are combined and solved to derive the
i1 {C1} and {C2}. And then the {U1}, {U2} and other variables such as
X  (15)
n
n5 for unidirectional infinite element the strain and stress can be calculated.
v Ni vi
i1
n3 for bidirectional infinite element In order to simulate the infinite domain of the asphalt
pavement, the infinite element was added to SAFEM first. The
where shape function Ni can be seen in Table 1. theoretical derivations and computational procedures of the
If only the finite elements are used, the calculation of stiffness matrix in finite element method and finite-infinite
stiffness matrix only involves the shape function. When the element coupling method are relatively similar, which is
infinite elements are applied, the calculation of their convenient for the application of the infinite elements. The
displacement still uses the shape function, while the coordi- global stiffness of the nodes at the interface can be calculated
nate transformation of the infinite elements must use the as the sum of the nodal stiffness of the finite and infinite
mapping function. element (Jiang et al., 2009).

Table 1 e Formula of mapping and shape function for the 2D mapping infinite element.
Infinite element type Mapping function Shape function
Unidirectional M1 (1  x)(1  x  h)/1  h N1 1/4(1  x)(1  h)(1  x  h)
M2 2(1  x2)/1  h N2 (1  x2)(1  h)
M3 (1 x)(1 x  h)/1  h N3 1/4(1 x)(1  h)(1 x  h)
M4 (1 x)(1 h)/2(1  h) N4 (1 x)(1  h2)
M5 (1  x)(1 h)/2(1  h) N5 (1  x)(1  h2)
Bidirectional M1 xh 3(1  x  h)/(1  x)(1  h) N1 1/4(1  x)(1  h)(1  x h)
M2 2(1 x)/(1  x)(1  h) N2 (1  x2)(1h)
M3 2(1 h)/(1  x)(1  h) N3 (1  x)(1  h2)
52 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8

2 3 2 3 2 3
/ / / / / / / / /
4 / kij / 5 4 / k1 / 5 4 / k2 / 5 (19)
ij ij
/ / / / / / / / /

where kij is the nodal stiffness in the global stiffness matrix, k1ij
and k2ij are the stiffness of node in the finite and infinite ele-
ments, respectively, i and j are the address indicators of the
node.

4. Verification of SAFEM by comparison with

ABAQUS

The accuracy of the SAFEM only with finite elements was

verified by comparison with the results derived from the
general commercial FE-program ABAQUS (ABAQUS, 2011).
The parameters and boundary conditions in the modeling of
these two programs were selected as consistent as possible to
ensure a high comparability of the models.

4.1. Definition of the models

The responses from both models were evaluated using the

pavement type in Table 2, which is widely used in Germany
according to the guidelines RStO 01 (FGSV, 2002) and RDO
Asphalt 09 (FGSV, 2009). The thicknesses of all layers
excepting the sub-grade were derived from RStO 01 (FGSV,
2002). The thickness of the sub-grade was defined as
2000 mm. Setting such a large value was aimed to minimize
the influence of the boundary condition on the results. Be-
sides, the length and width of all layers were set to 6000 mm
for the same reason. The pavement surface temperatures of
12.5  C (winter) and 27.5  C (summer) were assumed, and
then the associated material properties in the superstructure
were determined according to RDO Asphalt 09 (FGSV, 2009), as
listed in Fig. 4 and Table 2.
The mesh generator of SAFEM was used to create a 2D
mesh in the xy plane consisting of 6-node triangular elements,
as shown in Fig. 5(a). While the model in ABAQUS was three-
dimensional, so the type of 3D 10-node quadratic tetrahedron
element was applied in order to make the mesh geometry as
close as possible to that in SAFEM. The mesh in ABAQUS is
illustrated in Fig. 5(b). Due to the different dimensions of both
models, the mesh algorithms are different, but the element
size increases gradually from the loading area in both models.
Fig. 4 e Structure of the pavement.
The number of elements in SAFEM and ABAQUS are 2272 and
49,671, respectively.
According to RDO Asphalt 09 (FGSV, 2009), the load adopted
in SAFEM and ABAQUS was 49 kN circular load with the radius
of 150 mm, so the uniformly distributed contact stress was
0.7 MPa. The loading area was located at the center of the
pavement illustrated in Fig. 2.
The bottom nodes of the mesh representing the sub-grade
in both models were fixed in all directions. According to the
theory of SAFEM, the displacements on both edges (z 0 and
z a) are restricted to zero in the x- and y-directions. The
same boundary conditions were also used for the ABAQUS
model. The three asphalt layers are totally bound and the two
contact layers among the asphalt base course, road base
course, sub-base and sub-grade are defined as being partially
bound, which means the nodes at the interface between the
different layers always have the same displacements in the
vertical direction but may have different displacements in the
horizontal direction.

Table 2 e Geometrical data and material properties of the 4.2. Comparison of the results between ABAQUS and
pavement.
SAFEM
Layer m Winter Summer
E (MPa) E (MPa) The computed results from both models are compared in
Surface course 0.35 22,690 2902

pattern, as shown in Fig. 6. The cross-section is the
Moire
Binder course 0.35 27,283 6817 inner surface with x 3000 mm. It can be seen that the dis-
Asphalt base curse 0.35 17,853 4903 tribution of the stresses and the deformation shapes from
Road base course 0.25 10,000 10,000 both FE programs are consistent.
Sub-base 0.50 100 100 The computational stresses shown in Figs. 7e10 are
Sub-grade 0.50 45 45
derived from four series of response points offset from the
j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8 53

Fig. 5 e Mesh automatically generated from two FE-programs.

Fig. 6 e Computational stress and deformation from ABAQUS and SAFEM.

54 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8

Fig. 7 e Comparison of vertical stress on the top of asphalt surface course derived from ABAQUS and SAFEM.

Fig. 8 e Comparison of horizontal stress on the bottom of asphalt base course derived from ABAQUS and SAFEM.

loading center to the boundary (x coordinate from 3000 to

6000 mm in SAFEM of Fig. 6).
In Table 3, the vertical and horizontal stresses at four
critical points directly below the loading center, where
maximum compressive or tensile stresses may occur, are
considered. From Figs. 7e10, and Table 3, it can be stated that
the results from both programs have a high correlation except
for a slightly larger difference in the vertical stresses on the
top of the sub-base. The computation time of the SAFEM is
much shorter than that of the ABAQUS. Both FE analyses were
run on a computer with an Intel Core Duo 3.4 GHz, 32 GB RAM.
On average, the computational time required by the ABAQUS
3D model is about 220 s, whereas the SAFEM model requires
11 s. With code optimization, the computational time of the
SAFEM may be further reduced.
5. Verification of finite-infinite element
coupling analysis in SAFEM
5.1. Definition of the models
The reliability and efficiency of the finite-infinite element
coupling analysis in the SAFEM were studied by comparison of
the results with those from the SAFEM which have been
proved to be reliable in the section above. The asphalt pave-
ment structural responses from both analyses were evaluated
using the same pavement type with the material properties
shown in Table 2 (winter case).
The thicknesses of all layers excepting the sub-grade were
reference to Table 2. The thickness of the sub-grade was

Fig. 9 e Comparison of horizontal stress on the bottom of road base course derived from ABAQUS and SAFEM.
 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8 55

Fig. 10 e Comparison of vertical stress on the top of sub-grade derived from ABAQUS and SAFEM.

Table 3 e Comparison between ABAQUS and SAFEM regarding the determined stresses (MPa) at critical points.
Points Winter Summer
SAFEM ABAQUS Difference SAFEM ABAQUS Difference
1 0.704 0.704 0.000 (0%) 0.698 0.702 0.004 (0.57%)
2 0.938 0.959 0.021 (2.19%) 0.663 0.675 0.012 (1.78%)
3 0.262 0.269 0.007 (2.60%) 0.715 0.725 0.010 (1.38%)
4 0.010 0.009 0.001 (11.1%) 0.027 0.022 0.005 (22.7%)

defined from 100 to 9000 mm. Setting such a series of values is
aimed to investigate the influence of the sub-grade thickness
on the results. The length of each layer along traffic direction
was set from 3500 to 30,000 mm for the similar reason. The
width of the pavement was still set to 6000 mm.
The mesh generator of SAFEM was used to create a 2D
mesh consisting of 6-node triangular finite elements for the
finite element analysis (Persson and Strang, 2004) as shown in
Fig. 11(a). When coupled with infinite elements, the 5-node
unidirectional or 3-node bidirectional mapping infinite ele-
ments were created outside the finite element domain, as
shown in Fig. 11(b). If the geometrical parameters are the
same, the numbers of the finite elements are also the same in
both analyses, but the finite-infinite element coupling anal-
ysis has several additional infinite elements.
The load and the condition of interlayer connection were
defined as the same with that used in the Section 4.1. The
bottom nodes of the mesh representing the sub-grade in the
finite element analysis were fixed in all directions, but in the
finite-infinite element coupling analysis this boundary con-
dition was not necessary.
5.2. Comparison of the results between finite element
analysis and finite-infinite element coupling analysis
Four response points from the same critical points in Section
4.1 were selected to compare the results of both analyses.
If the pavement length is kept as 30,000 mm which is
considered to be large enough and hence does not influence
the computational results, a series of results are derived with
increasing the sub-grade thickness of the pavement, as shown
in Fig. 12.
When the results from both analyses come to convergence,
the absolute values of the results from the finite-infinite
element coupling analysis are a little larger than those from
the finite element analysis. All the relative errors of the four
response points (take the results of the finite element analysis
as reference) are below 1.3%, which proves the reliability of
the finite-infinite element coupling analysis in the SAFEM.
Furthermore, in the prediction of the pavement remaining
service life, the slightly larger stress values will make the
prediction result safer, which even has a certain positive
significance.

Fig. 11 e Mesh generated by SAFEM.

56 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8
Fig. 12 e Comparison of the results when the pavement length is kept constant.
The vertical stress on the top of the surface course does
not vary significantly, which means the thickness of the
sub-grade almost does not influence the stress on the
pavement surface. However, with increasing depth of the
response points in the pavement, the stresses vary more
significantly, especially on the top of the sub-base, i.e. the
influence of the sub-grade thickness becomes more and
more significant on stress state of deeper location in the
pavement.
Comparing the results between the two analyses, the in-
fluence of the sub-grade thickness on the results from finite
element analysis is much more significant. If its own
convergence value is taken as reference, all the relative errors
of the four response points reach the level below 3.5% in the
finite element analysis when the thickness of sub-grade
comes to be 3000 mm, while the thickness in finite-infinite
element coupling analysis only needs to be 1500 mm.
Similarly, when the thickness of sub-grade is kept as
7500 mm which does not influence the computational results,
a series of results from both analyses are derived with
increasing the pavement length, as shown in Fig. 13.
Except for the stress on the top of the surface course, the
stresses at the other response points vary similarly signifi-
cantly, which means the influence of the pavement length
on the stress is great regardless of the depth of the response
points. The convergence rate of the finite-infinite element
coupling analysis is still obviously higher than that of the
finite element analysis, i.e. when the pavement length rea-
ches 3500 mm in the finite-infinite element coupling analysis
the relative errors become below 3%, while the length in
finite element analysis needs to be 5000 mm to reach this
level.
In order to compare the computational accuracy and time
between the two analyses, two cases were studied as shown in
Table 4. The two analyses were run on a computer with an
Intel Core Duo 3.4 GHz, 32 GB RAM. If the minimum values of
sub-grade thickness and pavement length (Case 1) is applied
in previous investigation in both analyses, the accuracy of the
finite-infinite element coupling analysis still keeps a relatively
high level, but that of the finite element analysis decreases
extremely. Although the finite-infinite element coupling
analysis has additional nodes of the infinite elements, it does
not need to determine constrain condition on the boundary,
as a result the computational time (9 s) is even less than that
of the finite element analysis (11 s). In order to keep the rela-
tively high accuracy level in both analyses, the geometrical
parameters were increased to different levels as shown in
Case 2. In this case the computational time of the finite-
infinite element coupling analysis (14 s) is only 70% of that
of the finite element analysis (20 s), which reveals the higher
computational efficiency of the finite-infinite element
analysis.
6. Conclusions
This paper proposes to use the SAFEM for predicting the
asphalt pavement structural responses under static loads. A
computer program was developed based on MATLAB in which
the SAFEM was applied. The accuracy of the program is
 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8 57

Fig. 13 e Comparison of the results when the thickness of sub-grade is constant.

Table 4 e Comparison of the relative errors and computational times from both analyses.
Item Case 1 Case 2
Finite Infinite Finite Infinite
Thickness of sub-grade (mm) 100 100 3000 1500
Pavement length (mm) 3500 3500 5000 3500
Relative error (%) Top of the asphalt surface course 0.04 0.03 0.01 0.00
Bottom of the asphalt base course 25.04 4.49 3.58 3.54
Bottom of the road base course 24.21 1.89 3.48 3.44
Top of the sub-base 107.08 12.28 5.31 1.75
Computational time(s) 11 9 20 14

verified by a comparison with ABAQUS. Pavement responses the finite-infinite element coupling analysis should be deter-
to a static load predicted by SAFEM and ABAQUS are in very mined by the theoretical research and large numbers of case
good agreement. Furthermore, the computational time of the studies.
SAFEM is much shorter than that of the ABAQUS.
In order to further reduce the computational time, the
infinite elements are coupled with the finite elements in the
Acknowledgments
SAFEM. The investigation shows that the results of the finite-
infinite element coupling analysis are reliable and its
The work is part of a research project represented by German
convergence rate is much higher than that of the finite
Federal Highway Research Institute (BASt), financed by the
element analysis in the SAFEM. As a result, the scale of the
Federal Minister of Transport and Digital Infrastructure
pavement model at the infinite domain can be controlled in a
(BMVI), and conducted under FE 04.0259/2012/NGB.
suitable level and the computational time can be reduced
without decreasing its accuracy.
For further investigation, the SAFEM allows the application references
of dynamic analysis and various material properties, such as
viscoelasticity for asphalt and nonlinear elasticity for sub-
base of the pavement. Furthermore, the regulation of the ABAQUS, 2011. ABAQUS Analysis User's Manual. Dassault
minimum amount of the finite elements required to be used in Systemes Simulia Corp., Providence RI.
58 j o u r n a l o f t r a f fi c a n d t r a n s p o r t a t i o n e n g i n e e r i n g ( e n g l i s h e d i t i o n ) 2 0 1 5 ; 2 ( 1 ) : 4 8 e5 8
Beer, G., Meer, J.L., 1981. Infinite domain element. International
Journal for Numerical Methods in Engineering 17, 43e52.
Bettess, P., 1977. Infinite elements. International Journal for
Numerical Methods in Engineering 11, 53e64.
Bettess, P., 1980. More on infinite elements. International Journal
for Numerical Methods in Engineering 15, 1613e1626.
Bettess, P., Zienkiewicz, O.C., 1977. Diffraction and refraction of
surface waves using finite and infinite elements.
International Journal for Numerical Methods in Engineering
11, 1271e1290.
FGSV, 2002. Guidelines for the Standardization of the Upper
Structure of Traffic Areas (RStO 01), 2001 ed. FGSV Publisher,
Research Society for Road and Transportation, Cologne.
FGSV, 2009. Guidelines for the Computational Dimension of the
Upper Structure of Road with Asphalt Surface Course (RDO
Asphalt 09), 2009 ed. FGSV Publisher, Research Society for
Road and Transportation, Cologne.
Fritz, J.J., 2002. Flexible pavement response evaluation using the
semi-analytical finite element method. International Journal
of Materials and Pavement Design 3 (2), 211e225.
Hjelmstad, K.D., Zuo, Q., Kim, J., 1997. Elastic pavement analysis
using infinite elements. Transportation Research Board 1568,
72e76.
Hu, S., Hu, X., Zhou, F., 2008. Using semi-analytical finite element
method to evaluate stress intensity factors in pavement
structure. In: Proceedings of the 6th RILEM International
Conference on Cracking in Pavement, Chicago, 2008.
Jiang, Y., Chen, D., Yao, Y., 2009. Application of finite element and
infinite element method to stress analysis of gravity dam.
Engineering Journal of Wuhan University 42, 322e325.
Li, L., Kunimatsu, S., Wang, A., 2007a. Application of the
generalized infinite element method to unbounded elasto-
static problems. In: Proceedings of the ICCM2007, Hiroshima,
2007.
Li, L., Kunimatsu, S., Wang, A., 2007b. The infinite element
method and its application. Advances in Mechanics 37 (2),
161e174.
Liu, P., Chen, X., Oeser, M., et al., 2014a. Application of semi-
analytical finite element method in analysis of asphalt
pavement structural response. In: Proceedings of 12th
International Conference on Asphalt Pavements, Raleigh,
2014.
Liu, P., Wang, D., Oeser, M., 2013. Leistungsfa hige semi-analytische
methoden zur berechnung von Asphaltbefestigungen. In: 3rd
Dresdner Asphalttage, Dresden, 2013.
Liu, P., Wang, D., Oeser, M., et al., 2014b. Einsatz der semi-analytischen
finite-elemente-methode zur beanspruchungszusta nde von
asphaltbefestigungen. Bauingenieur 89 (7/8), 333e339.
Persson, P.O., Strang, G., 2004. A simple mesh generator in
MATLAB. Society for Industrial and Applied Mathematics 46
(2), 329e345.
Sallah, M., Buchanan, G.R., 1990. Some simplified methods for
infinite elements. Computational Mechanics 6, 167e172.
Ungless, R.F., 1973. An Infinite Finite Element. Master thesis.
University of British Columbia, Vancouver.
Wang, Q., Brill, D.R., 2013. Improvements in the application of
infinite elements to the NIKE3D program for airport
pavement response analysis. International Journal of
Pavement Engineering 14 (5), 429e439.
Zhao, B., 2012. Dynamic Analysis of Semi-infinite Soil Based on
Finite-element and Infinite-element Coupling Method.
Master thesis. Beijing Jiaotong University, Beijing.
Zhou, S., Hu, X., Wang, J., 2004. The application of infinite element
in numerical analysis of geotechnical engineering. Journal of
Chongqing Jiaotong University 23, 61e64.
Zhu, J., Bian, C., 2001. The analysis method and its application in
excavation which couples the finite element and infinite
element. Ground Improvement 12 (3), 17e22.
Zienkiewicz, O.C., Emson, C., Bettess, P., 1983. A novel boundary
infinite element. International Journal for Numerical
Methods in Engineering 19, 393e404.
Zienkiewicz, O.C., Taylor, R.L., 2005. The Finite Element Method
for Solid and Structural Mechanics, sixth ed. Elsevier
Butterworth-Heinemann, Oxford.